Re: RFS : ocamlgsl 0.3.2-2 and ocamlgraph 0.70-3 ( fix a RC bugs )
Hello,
On Mon, May 03, 2004 at 09:31:54PM -0400, Mike Furr wrote:
> per ocamlgsl,
>
> There's no reason to take sparc and hppa out of the arch list. The
> only problem is in the way ocamlgsl is using ocaml's foreign function
> interface(Double_val is actually a function, not a macro on these
> archs). So applying the attached 2 line patch will make it build.
> I didn't check this into svn, since I figured I let you do that
> with whatever style you prefer.
>
> Also, it would probably be a good idea to forward this upstream as
> well.
>
> After you check this in, I'd be happy to upload it for you.
>
> --
> Mike Furr <mfurr@debian.org>
> 1024D/124B26F3 5B9F 587F BC5C D823 50CE 4DB0 ED93 CA29 124B 26F3
> diff -urN ocamlgsl-0.3.2.orig/mlgsl_ieee.c ocamlgsl-0.3.2/mlgsl_ieee.c
> --- ocamlgsl-0.3.2.orig/mlgsl_ieee.c 2004-02-05 09:32:42.000000000 -0500
> +++ ocamlgsl-0.3.2/mlgsl_ieee.c 2004-05-03 21:34:46.000000000 -0400
> @@ -27,7 +27,8 @@
> value ml_gsl_ieee_double_to_rep(value x)
> {
> gsl_ieee_double_rep r;
> - gsl_ieee_double_to_rep(&(Double_val(x)), &r);
> + double d = Double_val(x);
> + gsl_ieee_double_to_rep(&d, &r);
> return rep_val(&r);
> }
>
;-)
Yep, i propose the same patch to upstream. He reply me with a simple
answer : "yep, it will compile, but it also will produce a bus error on
those architecture ".
In fact, for a lot of functions, it uses array of float, and when using
this you need to copy the full array casting each member of it.
In other words, upstream tell me that building for those arch require
too much work ! ( he just release a new tarball, which contains a #ifdef
ARCH_ALIGN_DOUBLE #error... ).
That is the reason why i exclude those arch from the build.
Kind regard
Sylvain Le Gall
Reply to: